Modular composition and state update in Plaid

Abstract

At the core of the Plaid typestate-oriented programming language is the ability to change the representation of an object at run-time. As such, the semantics of the state change operation impact how Plaid programs are structured and how objects are composed in Plaid's trait-based composition system. We describe the challenges with respect to designing a modular state change operation and suggest two options. Secondly, we explore the issues both designs create for explicit conflict resolution and sketch a potential solution which eliminates the flattening property to allow conflicts to be resolved only when they come into scope.

First-class state change in plaid

“…While we recap the motivation and concept of the language from this earlier work, that paper described an unimplemented language, and neither defined the language semantics nor investigated the modeling of complex state spaces, which are the key contributions of this paper. In an earlier 4-page workshop paper, we explored the need for a modular state change operator that affects only one dimension of state change at a time [2]; this paper gives the semantics for a concrete solution to that problem. Other recent work has begun to explore a gradual, permission-based type system for Plaid [29].…”

“…Thus, when we transition from an ClosedFile to an OpenFile the members of the File state should remain constant. This is the semantics behind the restricted update semantics of state change described in [2]. We use and extend these semantics in a natural way to account for our hierarchical object model.…”